Comment on page
🟣
Stacking and Training Strategies Playbook
One Life. Many Dimensions. Your Personal AI.
Life is continuous, so is your personal AI. We all have personalities, aspirations, and dimensions that define who we are and what we want to be. A Personal AI is designed to mimic the life experiences of a human being, to ambiently capture those moments happening around us, and to continuously learn from our past.
The idea of stacking and training isn’t a one-time thing. If you position yourself with an ‘always-learning’ mindset, your AI will help you reap the benefits. This isn’t a matter of training a model or bot just once; rather, a continuous process of learning and building the make-up of your mind. For instance, if one’s personal AI is being used strictly for specific use cases such as summarizing a document or creating a bot for an FAQ, it isn’t being used to its fullest extent.
It’s natural to jump into getting started by plugging all the data you have into your personal AI to extract value from it, but be sure to keep your AI updated on a regular basis to get the most out of it — developing into a true version of you.
You are investing in your AI model because it is yours to utilize to provide value to yourself and those around you. Don’t get deterred by the influence of pre-trained systems around you, as long as you maintain unique knowledge and expertise — your personalized AI system will bring unique value to you.
Personal AI is a messaging app in which your AI acts as an extension of you. Your AI learns from messages you send to it and others, as well as from your selections of the suggested responses it provides you.
The UX of your personal.ai app will be similar to that of iMessage, Whatsapp, or Slack.
A message can be just text, an image, a web link, a YouTube clip, rich text, or a file. A message can be sent privately in DM to yourself (your AI) or to others. An incoming message is interpreted as context, intent, or prompt to your AI to draft a response for based on your chosen co-pilot or auto-pilot settings.
Stacking- the process of capturing and organizing raw data for your AI.
Training- the process of using algorithms to create a personal language model off your stack. (This is done automatically at the moment with additional capabilites to be made available soon.)
AI Profiles- subsets of data created to organize data within given scopes, to then be able to control which subsets you’d like to share where.
You stack by sending messages to yourself (your AI).
Most people associate sending text messages to themselves as their favorite way of keeping track of information on the go. Your personal AI is built on this logic to serve this purpose for you. Any message you send to yourself in app will be remembered by your AI.
In contrast to many messaging services where the intention of sending links and files is to share them with other human beings, your personal AI instead attempts to extract the data from the link or files and stack it.
Here’s a breakdown of how various types of Messages are interpreted by your AI:
- Short Text or question -> not stacked but consulted with your personal AI for a reply.

Question to my AI (inquiry) with response
- Longer Text (sentence, paragraph) -> stacked and consulted with your personal AI for a reply.

Paragraph that is stacked
- Link -> if accessible, (websites, articles, connected Google Docs) the content is stacked.

URL to my AI
- Image -> stacked and consulted with your personal AI for a reply (experimental).
- YouTube video link -> transcribed and stacked.

Youtube video to my AI
- File attachments -> stacked (coming soon in 2.0 — support for docs, pdfs, and audio files).
- Clipboard -> the same above rules apply for text coming into your AI via the sync clipboard integration when it is turned on.
You stack by messaging and by responding to messages — with your friends, coworkers, and wider community.
With the 2.0 release, Personal AI will be a messaging tool where every message you send to someone will be used to train your AI. Your AI will not only learn from those messages but also will start drafting suggestions from your own AI.
You stack by reusing old data sources that you have accumulated over time (Twitter, Clipboard, Google Docs, Google Calendar, and Gmail).
Use integrations to speed up your AI’s learning. Press Cmd+K -> Click Integrations to explore. There are two types of integrations associated with each data source:
- 1.Connect integration- which will load a URL one at a time (such as with GDrive currently).
- 2.Sync integration- where the data automatically syncs once connected.

Memory blocks in memory stack via Twitter integration
- 1.Automated base training (available for everyone): As your Stack grows, the base model is phased out and continues to train and evolve. Given the facts and experience, the base model is more like a baby model that answers questions and generates some thoughts given the message prompts. It’s a personal question-and-answer model that uses shared computing resources.
- 2.On-demand dedicated training (coming soon): The base model has some limitations such as the number of inference calls (interactions with the AI), the output length of 256 tokens (2 paragraphs), and latency (time for your AI to respond to multiple requests). The advanced model is a dedicated model that runs in its own container. This model will eventually be run on edge such as mobile or computer. It can support up to 100 people talking to your AI and for now, has no limits on inference calls (interactions with the AI).
- 3.On-demand conversational training (on the roadmap): Using the “reply” function, the base and advanced models have memories of what was said in the previous message. To get back-and-forth messages in a discussion, your personal AI would be required to recall what was said for the last few messages, not only the most recent one. The conversation model will have a long-term memory (the ability to remember the context of many turns on what people are saying back and forth) and will be able to provide further suggestions.
The primary purpose of your personal AI is to foster deeper conversations with coworkers and peers by using co-pilot and auto-pilot modes that focus on your true forms of expressions when they are available.
You’ll now get to decide what kind of discussions you want to have in your personal AI and with whom:
Use Personal AI Profiles if you have multiple dimensions of what you want to train within your multi-faceted life.

Framework for setting purposes for your personal AI.
- 1.Message in DMs or Lounges with My Close F&F about My Life trained in an AI profile to Deepen Connections.
- 2.Train this AI profile on personal facts, preferences, stories, hobbies, traits, etc that are usually captured by journaling or sharing them in text chats and conversations. Integrate into your daily life by messaging text with your F&F or yourself and fast track historical data input by importing from iMessage, Whatsapp, and Gmail (coming soon).
- 3.Stack data by formatting this information as if you are an autobiographer in the first person. Here is an example of your outgoing messaging to yourself or others: “I was born in India. My favorite color is yellow. My co-founder is @Sharon Zhang”. Then your personal AI can draft responses for an incoming message such as: “I wanted to make you a personalized gift for your birthday! What color do you like the most?”

Direct message with AI response from previously stacked message
- 1.Message in DM or Lounges with My Social Network about My Activities trained in an AI profile to Stay Informed.
- 2.Train this AI profile on personal posts, announcements, opinions, events, etc that are usually captured in social media posts, blogs, speeches, and calendars. Integrate into your daily life by messaging share links of social media posts or copy and pasting data and fast track historical data input by importing from Twitter, Youtube, Medium (available) and Calendar, Facebook, LinkedIn (coming soon).
- 3.Stack data by formatting this information as if you are an speaker on stage in the first person with clear tone expression. Here is an example of your outgoing messaging to yourself or others: “I think we can all argue that humans are now more selfish than ever before with little to no empathy for others.”. Then your personal AI can draft responses for an incoming message such as: “Dude…the people in that meeting were ridiculously rude.”

Direct message with AI response from previously stacked message
- 1.Message in DM or Lounges with My Team about My Projects trained in an AI profile to Collaborate Effectively.
- 2.Train this AI profile on personal information and data exchange in settins such as a group, company, community org etc in a collaboration setting that are usually captured by communicating, collaborating and sharing information. Integrate into your daily life by messaging in AI-lounges designated for the project purpose and using text, URLs and attachments for exchanging information and fast track historical data input by importing from Drive, Youtube (available) and Slack, Discord, Dropbox and OneDrive (on roadmap).
- 3.Stack data by formatting this information as if you are an manager communicating information in first person and full statements. Here is an example of your outgoing messaging to yourself or others: “A maintenance banner inside the app might be the most helpful for the weekend cutover.”. Then your personal AI can draft responses for an incoming message such as: “any ideas how you want to tell customers about this weekend’s maintenance?”

Direct message with AI response from previously stacked message
- 1.Message in DM or Lounges with My Customers and Clients about My Products and Services trained in an AI profile to Save Time Responding.
- 2.Train this AI profile on personal content useful in specific context such as how-to, books, newsletters, faqs, material etc that are usually captured by documenting, emailing, creating websites and videos. Integrate into your daily life by messaging attachments, supported links and copy pasting content and fast track historical data input by importing from Youtube, Drive (available), Gmail (coming soon) and Crawler, Coursera, Udemy (on roadmap).
- 3.Stack data by formatting this information as if you are an specialist quickly getting right information across. Here is an example of your outgoing messaging to yourself or others: “To upgrade to the community plan, email [email protected] with information about your use case and community size.” Then your personal AI can draft responses for an incoming message such as: “how can I upgrade to the community plan?”

Direct message with AI response from previously stacked message
- 1.Message in DM or Lounges with My Self about My Knowledge trained in an AI profile to Create Second Brain.
- 2.Train this AI profile on personal knowledge from research, reading, meeting, listening and watching that are usually captured in pdfs, blogs, note taking tools, meeting recording tools, books, podcasts, videos and tv shows. Integrate into your daily life by messaging notes, attaching files, supported links, and copy pasting data and fast track historical data input by importing from RSS Feeds, Obsidian/Roam, Zoom, Podcasts Lists (on roadmap, use Zapier meanwhile).
- 3.Stack data by formatting this information as if you are a journalist with who said what and citations. Here is an example of your outgoing messaging to yourself or others: “#Blockchain is useful for #AI in applications such as data #ownership. As @Forte Tiago said network of data is all that is required for creating a second brain. I can group all relevant information in :Second Brain Info”. Then your personal AI can draft responses for an incoming message such as: “how are #Blockchain and #AI related to each other?”

AI question and response leveraging topics (#) to specify data query
Set your AI purpose. Choose your messaging needs. Remember to Never Forget.
Have Fun!